 |
 |
Tuesday, Nov 24
|
|
|
|
OVERALL DESIGN
The PSID gathers information about families and all individuals
in those families through its annual interviews.1 A single
primary adult--usually the male adult head,2 if there is
one--serves as the sole respondent. Sometimes the wife (or
cohabitor, referred to as "wife") of the head agrees to grant an
interview when the head does not. The single household respondent
provides information about him/herself and about all other family
members.3
The study's original households constitute a national probability
sample of U.S. households as of 1967. Its rules for following
household members were designed to maintain a representative
sample of families at any point in time as well as across time.
To accomplish this, the PSID tracks members of its wave-1 (1968)
families, including all those leaving to establish separate
family units. Children born to a member of an original-sample
member are classified as sample members and are eligible for
tracking as separate family units when they set up their own
households. Ex-spouses and other adult sample members who move
out of PSID family units are tracked to their new family units.
This procedure replicates the population's family-building
activity and produces a dynamic sample of families each year. New
PSID families form when children grow up and establish separate
households or when marriage partners go separate ways. This
results in growth over time in both the number of family units
and the number of people residing with a sample member at some
time during the study.
Information is gathered about all persons residing in the family
unit, but in most waves there is only one respondent per family
unit (usually the head). The most detailed information is
collected each year about the heads of family units. Since the
late 1970s, however, the PSID has sought to collect the same
detail for wives/"wives" (by "wives" we mean cohabitors) as for
heads. For special supplements gathering retrospective history
information in 1976 and 1985, the study conducted separate
interviews with all wives/"wives" of heads as well as their
husbands. Except for the very early years of the study,
cohabitors have been treated in a similar manner to husbands and
wives.
The general design of the study has remained largely unchanged
over time; however, the mode of interviewing has changed. From
1968 through 1972, the PSID conducted in-person interviews. In
1973, to reduce costs, the study began taking the majority of
interviews by telephone. Since that time, in-person interviews
have been conducted only with respondents who do not have
telephones (roughly 500 each year), or who have special
circumstances which make a telephone interview unfeasible. To
further reduce costs, and because long interviews are difficult
by telephone, interview length was also reduced in 1973. The
interview averaged about one hour when it was conducted in
person; since the change to telephone interviewing the length has
averaged 20 to 30 minutes.
As discussed in the "Content" chapter, the PSID has maintained a
core of questions addressing issues relevant to income dynamics
and demographic change. In addition to the central core, there
have been a number of supplements to the core, adding questions
on a wide variety of other topics. These supplements have led to
the creation and release of a number of special files that
complement the main PSID data files.
SAMPLE DESIGN
Sample frame.
The initial sample for the PSID actually consisted of two
independent samples: a cross-sectional, national sample (based
on stratified multistage selection of the civilian
noninstitutional population of the U.S.) and a national sample of
low-income families.4 The cross-section sample was drawn by the
Survey Research Center (SRC). Commonly called the SRC sample, it
was an equal probability sample of households in the 48
coterminous states designed to yield about 3,000 completed
interviews. (In fact 2,930 interviews were taken in 1968 from
this sample).
The second sample of responding PSID families, known as the SEO
sample, came from the Survey of Economic Opportunity (SEO),
conducted by the Bureau of the Census for the Office of Economic
Opportunity. The PSID selected from the SEO's sample, the goal
being to obtain about 2,000 low-income families with heads under
60 years old. In fact, 1,872 families were successfully
interviewed. The SEO sample was confined to Standard Metropolitan
Statistical Areas (SMSAs) and to non-SMSAs in the Southern
region, and it involves unequal selection probabilities.
Both the SRC and SEO were subject to nonresponse in the first
wave (1968). Three factors played a special role in preventing
successful interviews with the SEO sample:
- There was nonresponse in the original Census survey from
which the SEO sample was selected.
- Sampled Census respondents were asked by the Bureau of the Census
to sign a release to allow their names to be passed on to SRC.
Approximately one quarter of the households failed to sign the
release.
- OEO failed to transmit some sampled addresses to SRC.
The PSID sample combines the SRC and SEO samples. Both samples
are probability samples (i.e., samples for which every element in
the population has a known non-zero chance of selection). Their
combination is also a probability sample. The combination,
however, is a sample with unequal selection probabilities, and as
a result compensatory weighting is needed in estimation, at least
for descriptive statistics. (The various disciplines disagree
about the need for weighting in model-based estimation.) Weight
adjustments are also needed to attempt to compensate for
differential nonresponse in 1968 and subsequent waves. As
explained in the "Data Analysis" chapter, and detailed in the
PSID's technical documentation, weights supplied on PSID data
files are designed to compensate for both unequal selection
probabilities and differential attrition.
Latino supplemental sample.
The original PSID sample contains too few Latino households to
provide reliable estimates either for Latinos as a group or for
major subgroups of Latinos. In addition, Latinos entering the
U.S. since 1968 are not represented in the basic PSID sample
unless they co-reside with persons in the U.S. in 1968. To help
reduce these shortfalls a sample of 2,043 Latino households was
interviewed and added to the PSID sample beginning in the 1990
wave. Funding for this supplemental sample came from the Ford
Foundation, the Rockefeller Foundation, the ASPE in the
Department of Health and Human Services, the Employment Training
Administration in the Department of Labor, and NSF.
The 1990 addition of a Latino sample is designed to provide
precisely the kind of representative information about Latinos
that is now available for blacks and non-Latino whites in the
original PSID. The sample was originally selected for the Latino
National Political Survey (LNPS), a 1989 study of the political
participation of Latino households cosponsored by the
InterUniversity Program for Latino Research and administered by
the Center for Mexican American Studies of the University of
Texas at Austin. The LNPS did not attempt to cover the entire
Latino population in the U.S. However, it covered at least 89
percent of the three largest Latino subgroups--the Mexican,
Puerto Rican and Cuban origin populations.
Latino sample members were asked extensive background information
in their initial interview in 1990, including marital and
fertility histories. In addition, questions were added to the
1990 interview to enable calculation of selection-probability
weights so that the Latino sample can be combined with the
existing PSID sample for analysis purposes. The PSID plans to
continue gathering comparable information from the Latino and
original PSID samples and to combine the two samples in PSID data
files.
Tracking rules.
The PSID's tracking rules call for following members of the
original family units and their adult offspring to whatever
living arrangements they experience. Information is gathered
about these sample members and their co-residents if they are
living in a household (i.e., non-institutional) situation. A
family member who moves out of a PSID family is eligible for
interviewing as a separate family unit if he or she is a sample
member and he or she is 18 years old or older and living in a
different, independent household.5 If a sample member 18 or older
moves to an institution such as a prison, a college dormitory, or
the military the PSID records this fact and attaches an
institutional status data record to the family he or she left.
The PSID keeps track of the location of sample members living in
institutional housing. Interviews are attempted with them if and
when they leave the institution to set up their own households.
| |
Institute for Social Research |
University of Michigan |
Privacy |
Conditions of Use
|
